A Laser Range Scanner Designed for Minimum Calibration Complexity

James Davis, Xing ChenJames Davis, Xing ChenStanford Computer Graphics LaboratoryStanford Computer Graphics Laboratory

3D Digital Imaging and Modeling3D Digital Imaging and Modeling

3DIM 20013DIM 2001

2

Scanner Designs

3

Scanner Design A

Quality Tradeoff

• Expense vs. accuracyExpense vs. accuracy

• Different curves possibleDifferent curves possible

• Complex calibration is expensiveComplex calibration is expensive

Expense

Acc

urac

y

Scanner Design B

4

Conventional stripe scanner

• Triangulation between camera and laserTriangulation between camera and laser

5

Complexities of traditional design

• Actuated componentsActuated components

• Cylindrical lens precisionCylindrical lens precision

• Custom calibration procedureCustom calibration procedure

6

Our design

• Triangulation between two camerasTriangulation between two cameras

• No actuated components No actuated components

7

Catadioptric layout

8

Our scanner

• Simple componentsSimple components• Camcorder, four mirrors, rigid mounting

9

Stripe processing

• Locate corresponding pointsLocate corresponding points• Use epipolar constraint

• Discard ambiguous data

10

Cylindrical lens precision

• No precision mountingNo precision mounting

                                 

    

11

Scanner calibration

• Camera model and pose Camera model and pose [Heikkila, Silven 97] [Heikkila, Silven 97]

• Well-studied easy calibrationWell-studied easy calibration

12

• Filter imageFilter image• Video signal noise

• Sub-pixel detectionSub-pixel detection• Local Gaussian approximation

Peak detection

Maximum intensity Local Gaussian Gaussian with filtering

13

Mesh coverage

• Video sequence defines meshVideo sequence defines mesh

• Stripe spacing related to laser velocityStripe spacing related to laser velocity

14

Constructing a mesh

• Fill stripe sampling gapsFill stripe sampling gaps

• Detect depth discontinuitiesDetect depth discontinuities

15

Aligning scans

• Iterative closest point (ICP) Iterative closest point (ICP) [Besl, McKay - Chen, Medioni 92][Besl, McKay - Chen, Medioni 92]

• Global alignment Global alignment [Pulli 99][Pulli 99]

16

Merging scans

• Volumetric merging (VRIP) Volumetric merging (VRIP) [Curless, Levoy 96][Curless, Levoy 96]

• Hole fillingHole filling

17

Depth resolution

440 mm x 550 mm

240x240 pixels

Expected depth resolution: 1.8 mm

18

Conclusion

• Minimal calibration complexityMinimal calibration complexity• No actuated components

• No precision lens placement

• Well-studied easy calibration model